# Copyright (C) 2022 ISIS Rutherford Appleton Laboratory UKRI
# SPDX - License - Identifier: GPL-3.0-or-later
from functools import partial
import numpy as np
import scipy.ndimage as scipy_ndimage
from mantidimaging.core.data import Images
from mantidimaging.core.operations.base_filter import BaseFilter, FilterGroup
from mantidimaging.core.parallel import shared as ps
from mantidimaging.core.utility.progress_reporting import Progress
from mantidimaging.gui.utility import add_property_to_form
from mantidimaging.gui.utility.qt_helpers import Type
OUTLIERS_DARK = 'dark'
OUTLIERS_BRIGHT = 'bright'
_default_radius = 3
_default_mode = OUTLIERS_BRIGHT
DIM_2D = "2D"
DIM_1D = "1D"
[docs]
class OutliersFilter(BaseFilter):
"""Removes pixel values that are found to be outliers as defined by the given parameters.
Intended to be used on: Projections
When: As a pre-processing step to reduce very bright or dead pixels in the data.
Caution: This should usually be one of the first steps applied to the data, flat and dark
images, to remove pixels with very large values that will cause issues for flat-fielding.
"""
filter_name = "Remove Outliers"
link_histograms = True
@staticmethod
def _execute(data, diff, radius, mode):
# Adapted from tomopy source
median = scipy_ndimage.median_filter(data, radius)
if mode == OUTLIERS_BRIGHT:
return np.where((data - median) > diff, median, data)
else:
return np.where((median - data) > diff, median, data)
[docs]
@staticmethod
def filter_func(images: Images,
diff=None,
radius=_default_radius,
mode=_default_mode,
cores=None,
progress: Progress = None):
"""
:param images: Input data
:param diff: Pixel value difference above which to crop bright pixels
:param radius: Size of the median filter to apply
:param mode: Whether to remove bright or dark outliers
One of [OUTLIERS_BRIGHT, OUTLIERS_DARK]
:param cores: The number of cores that will be used to process the data.
:return: The processed 3D numpy.ndarray
"""
if not diff or not diff > 0:
raise ValueError(f'diff parameter must be greater than 0. Value provided was {diff}')
if not radius or not radius > 0:
raise ValueError(f'radius parameter must be greater than 0. Value provided was {radius}')
func = ps.create_partial(OutliersFilter._execute, ps.return_to_self, diff=diff, radius=radius, mode=mode)
ps.shared_list = [images.data]
ps.execute(func,
images.num_projections,
progress=progress,
msg=f"Outliers with threshold {diff} and kernel {radius}")
return images
[docs]
@staticmethod
def register_gui(form, on_change, view):
_, diff_field = add_property_to_form('Difference',
'float',
1000,
valid_values=(1e-7, 10000),
form=form,
on_change=on_change,
tooltip="Difference between pixels that will be used to spot outliers.\n"
"It is calculated by subtracting the original image "
"from the median filtered image")
diff_field.setDecimals(7)
_, size_field = add_property_to_form('Median kernel',
Type.INT,
3, (1, 1000),
form=form,
on_change=on_change,
tooltip="The size of the median filter kernel used to find outliers.")
_, mode_field = add_property_to_form('Mode',
Type.CHOICE,
valid_values=modes(),
form=form,
on_change=on_change,
tooltip="Whether to remove bright or dark outliers")
return {'diff_field': diff_field, 'size_field': size_field, 'mode_field': mode_field}
[docs]
@staticmethod
def execute_wrapper(diff_field=None, size_field=None, mode_field=None):
return partial(OutliersFilter.filter_func,
diff=diff_field.value(),
radius=size_field.value(),
mode=mode_field.currentText())
[docs]
@staticmethod
def group_name() -> FilterGroup:
return FilterGroup.Basic
[docs]
def modes():
return [OUTLIERS_BRIGHT, OUTLIERS_DARK]